NGMN examines the AI surge and what it means for 6G

The Next Generation Mobile Networks Alliance (NGMN Alliance e.V.) has published a timely new whitepaper, AI Surge and its Implications for 6G, setting out an operator-led view on how artificial intelligence should be addressed as 6G standardisation gathers pace.

The backdrop is clear. AI is advancing at extraordinary speed. Large language models, multi-modal systems and increasingly autonomous AI agents are beginning to influence how applications are built, how devices behave and how services are consumed. Yet while AI’s societal impact is already visible, its precise impact on mobile networks remains uncertain. That uncertainty is at the heart of NGMN’s message: flexibility must be a core design principle for 6G.

The paper consolidates the perspectives of multiple mobile network operators and structures the discussion around three dimensions: the impact of AI-driven traffic on networks, the concept of the network for AI, and AI for the network including its architectural implications.

On traffic, the report takes a balanced view. Today’s mobile networks are still dominated by video consumption, and most AI interactions remain text-based. However, future scenarios could look very different. Multi-modal AI services, AI-enabled wearables such as AR glasses, autonomous vehicles, drones and industrial robots all have the potential to reshape traffic patterns. In particular, uplink demand could increase significantly if devices continuously capture and transmit sensor data, images or video for AI processing.

At the same time, several countervailing forces exist. AI models are being optimised for on-device inference. Chipsets are becoming more capable. Privacy and regulatory constraints may limit always-on sensing use cases. Some AI workloads may remain largely cloud-based. The conclusion is not that traffic will definitely explode, but that it might. As a result, 6G should be designed to adapt to changing uplink and downlink ratios, localised traffic surges and new forms of machine-oriented communication without requiring disruptive architectural redesign.

The second dimension, network for AI, moves beyond pure connectivity. NGMN argues that 6G must provide capabilities tailored to AI-native services. This includes more advanced and granular Quality of Service handling, better policy control, distributed edge computing support and mechanisms for explicit expression of an AI application’s networking and computing needs.

Importantly, the paper highlights the difference between theoretical performance targets and actual business value. For many current generative AI services, latency is dominated by compute processing time rather than transport delay. In such cases, simply pushing network latency ever lower may not translate into meaningful user experience gains. Enhancements should therefore be justified by real service requirements and value creation, not by technology ambition alone.

Beyond performance, the report explores new charging concepts such as token-based models reflecting fine-grained resource usage, dynamic and intent-driven networking for groups of collaborating AI agents, unified data and model frameworks across domains, and robust trust and authentication mechanisms. As AI agents begin to interact with networks and with each other, digital identity, authorisation and security frameworks will become critical.

The third dimension, AI for the network, looks inward. AI is expected to play a major role in 6G network management, orchestration and optimisation. Intent-driven management, autonomous operation, intelligent cross-domain orchestration and energy optimisation are all seen as important evolution paths. However, NGMN is careful to stress that AI should be applied where it brings clear value. In domains such as the radio access network, AI may deliver strong benefits in some layers while offering limited gains in others. Real-world validation, energy impact assessment and lifecycle management are essential.

Architecturally, the message is evolutionary rather than revolutionary. 6G is not envisaged as a clean-slate redesign. The 5G Service-Based Architecture is expected to serve as the starting point for the 6G core, with enhancements introduced where justified by AI-driven use cases. Concepts such as AI agent frameworks in the core, agent-to-agent and agent-to-network communication, and possible adoption of protocols like Model Context Protocol are discussed in the context of interoperability and multi-vendor environments.

The paper also highlights several risks. AI capabilities are evolving extremely quickly, raising the possibility that prematurely standardised features could become obsolete by the time 6G is deployed. There are also tensions between adding AI functionality and maintaining architectural simplicity, cost control and sustainability goals. Responsible AI principles, regulatory compliance and continued support for non-AI alternatives where required are all part of the broader picture.

For standards bodies such as 3GPP, NGMN recommends focusing on adaptable architectures, explicit support for AI service demands in terms of QoS and computing, interoperability and trust frameworks for multi-vendor and multi-agent environments, and mechanisms to support agent-to-agent and agent-to-network communications.

Overall, this publication reinforces a consistent theme in NGMN’s recent 6G work. 6G should be flexible, sustainable and value-driven. AI will be central to its evolution, but it should not dictate a disruptive reset of the mobile system. Instead, the industry is encouraged to integrate AI pragmatically, balancing innovation with operational reality and long-term investment protection.

As 6G studies progress, the real challenge will be managing uncertainty. AI may reshape traffic patterns, service models and network operations in profound ways. Or it may evolve in directions that reduce network dependency through greater on-device intelligence. Designing 6G to accommodate both possibilities is arguably the most important takeaway from this white paper.

Related Posts: 

Posted by on

Comments

Post a Comment